A method is suggested to ensure independence between the estimated global flow and changes in local flow in PET activation studies. Global flow is estimated as the average of all intracerebral voxels, except those that exhibit a consistent change in flow as a consequence of the experimental design. This is achieved by performing an initial analysis, using all intracerebral voxels for estimation of global flow, in which an F-map depicting task related changes is produced. A second analysis is performed, now excluding all voxels with a P < 0.05 in the F-map when evaluating global flow, thereby ensuring independence from areas changing in a task-dependent manner. The feasibility of the method is demonstrated on phantom data, showing that the distribution is skewed in nonactivated areas when including the activated areas in the calculation of global flow. By excluding these areas the distribution is translated toward zero and becomes consistent with the null-hypothesis. Furthermore, the usefulness of the suggested scheme is demonstrated on human data. Scans performed while subjects watched a movie featuring snakes were contrasted to scans performed while watching white noise, producing highly significant activations of the visual system. When using the traditional way of estimating global flow, deactivations were observed in very large portions of the frontal, parietal, and temporal lobes. When using the method suggested in the present paper these disappeared, which would be more consistent with the expected effects of stimulation. The method was used in conjunction with both ratio and ANCOVA adjustment with very similar results.
Copyright 1997 Academic Press.